logo separator

[mkgmap-dev] max-jobs patch

From Gerd Petermann gpetermann_muenchen at hotmail.com on Thu Feb 8 08:38:33 GMT 2018

Hi Manfred,

thanks for testing. Just to make this clear: I always use --max-jobs, because I also
always use -Xmx whenever I compile many tiles.

In this case a large tile is one that results in a large *.img
(one with many MB). Typically this depends on the --max-nodes option for splitter
and the style and the mkgmap options (routable or not, without or without DEM etc).

The max-jobs option allows to specify the number of parallel jobs,
e.g. --max-jobs=5 would start 5 parallel jobs (threads). Whenever one input file is done,
the next one is started until all are finished. The patch only changes the default from 1
to "number of cores". You can also use max-jobs=10 on a 2-core cpu, but this will slow down
the run time.

What I tried to point out is that a user who doesn't know about the --max--jobs option
might as well not know about the meaning of the -Xmx JRE option.

What happens if you start mkgmap on an 8-core machine with max-jobs=8 and
without -Xmx option? I think it will crash, and with the 2nd version of the patch it will tell
the user to look at the -Xmx option. So far so good.
What happens on a 4 core machine with max-jobs=4 and no -Xmx option?
What if the OS is 32 bit only? This would not allow much more than -Xmx1700m.
What if the option is used but set to a rather low value, e.g. -Xmx=1000m?

My current thinking is this:
If --max-jobs is not specified mkgmap should check the number of cores and the available heap
and assume that one thread needs ~500M. 

So, I would use something like num = Math.min(Runtime.getRuntime().maxMemory() / (512 * 1024 * 1024),
 Runtime.getRuntime().availableProcessors()) as default, maybe also for the case that one
specifies --max-jobs .

How does that sound?

Gerd






________________________________________
Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Manfred Haiduk <mhaiduk at t-online.de>
Gesendet: Mittwoch, 7. Februar 2018 22:28
An: mkgmap-dev at lists.mkgmap.org.uk
Betreff: Re: [mkgmap-dev] max-jobs patch

I don't know, if Nordrhein-Westfalen is that kind of large tiles you
meant. Splitter ended up with 43 files and i tried to make mapsource
maps with and without --max-jobs option and mkgmap r4000. My computer
does the job with 43 files about 12minutes faster than without this
option. My laptop is a core i5 with 8Gb of memory. The command i use is
like this line :

java -Xmx7500m -XX:-UseGCOverheadLimit -ea -jar
-Dlog.config=logging.properties
C:\PROGRA~2\KartenTools\mkgmap\mkgmap.jar
--location-autofill=is_in,nearest --mapname=70060001 --family-id=7006
.\styles\mystyle.txt --series-name="STYLE TEST"
--family-name="OpenStreetmap" --country-name=STYLETEST
--country-abbr=STY --area-name=STY --overview-mapname="overview"
--latin1 --style-file=.\styles --style=myland --max-jobs --keep-going
--check-roundabouts --drive-on=detect,right --output-dir=.\out\MYSTYLE
--index --bounds=mybounds --route
--name-tag-list=name,place_name,loc_name --housenumbers
--x-split-name-index --add-pois-to-areas --precomp-sea=.\input\sea.zip
--tdbfile --draw-priority=10 --gmapsupp -c .\data\MYSTYLE\template.args
--description=MYSTYLE --remove-ovm-work-files=true

Hope this gave you some useful hints


Am 07.02.2018 um 20:06 schrieb Gerd Petermann:
> Hi Mike,
>
> I did not yet try it. My understanding is that the Garbage Collection will try hard
> to avoid an out of heap exception. In other words: A user might see better throughput
> with fewer parallel jobs. What are your experiences? Did you try with
> large maps, say > 40 large tiles ?
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Mike Baggaley <mike at tvage.co.uk>
> Gesendet: Mittwoch, 7. Februar 2018 01:58
> An: 'Development list for mkgmap'
> Betreff: Re: [mkgmap-dev] max-jobs patch
>
> Hi Gerd, please find attached v2 of the max-jobs patch.
>
> This version additionally catches out of heap memory exceptions and displays
> a message suggesting either the use of the Java -Xmx option to increase the
> available heap memory or the mkgmap --max-jobs option with a smaller value
> to reduce the memory requirement (the latter is suggested only if using more
> than one thread).
>
> How does that seem?
> Regards,
> Mike
>
> -----Original Message-----
> From: Gerd Petermann [mailto:gpetermann_muenchen at hotmail.com]
> Sent: 05 February 2018 07:28
> To: Mike Baggaley <mike at tvage.co.uk>; 'Development list for mkgmap'
> <mkgmap-dev at lists.mkgmap.org.uk>
> Subject: Re: [mkgmap-dev] max-jobs patch
>
> Hi Mike,
>
> thought about this again. Maybe this change is too simple. With multiple
> jobs one also
> needs more heap (-Xmx JRE option). If you create rather large tiles with
> splitter (max-nodes=1600000)
> you need 0.5 - 1 GB for each job. Not sure what happens when a user creates
> a map with
> 10 tiles on an 8-core machine without any -Xmx option.
>
> I fear this will result in OutOfMemory exception, so better check the
> available heap as well.
>
> Gerd
>
> ________________________________________
> Von: mkgmap-dev <mkgmap-dev-bounces at lists.mkgmap.org.uk> im Auftrag von Mike
> Baggaley <mike at tvage.co.uk>
> Gesendet: Sonntag, 4. Februar 2018 15:14
> An: 'Development list for mkgmap'
> Betreff: [mkgmap-dev] max-jobs patch
>
> Hi Gerd,
>
> Please find attached a patch that amends the default behaviour if the
> --max-jobs option is not specified, to using a value equal to the number of
> CPU cores, as suggested in a previous post. The documentation is also
> amended to reflect the change. This halves the execution time of mkgmap for
> building a map of Staffordshire on my 8-core PC when --max-jobs is not
> specified (I didn't know about this option previously and was unaware the
> performance could have been improved).
>
> Cheers,
> Mike
>
> _______________________________________________
> mkgmap-dev mailing list
> mkgmap-dev at lists.mkgmap.org.uk
> http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev
>
--

  Mit freundlichen Grüßen

#####################################################
Manfred Haiduk,
Zum Fischbach 9,
52393 Hürtgenwald
e-mail mhaiduk at t-online.de
#####################################################

_______________________________________________
mkgmap-dev mailing list
mkgmap-dev at lists.mkgmap.org.uk
http://www.mkgmap.org.uk/mailman/listinfo/mkgmap-dev


More information about the mkgmap-dev mailing list